Search

Spin-transfer torque magnetoresistance random access memory is a major contender for static random access memory replacement in embedded caches at advanced fin field effect transistor nodes. It suffers, however, from the low resistance difference between the bistable states of the magnetic tunnel junction (MTJ). Variability on MTJ resistance and access transistors makes reliable read-out even more challenging. This triggered the use of complementary cells for low level caches needing high performance. This paper, focusing on the lower level caches, shows an improved 3T 2MTJ cell with a ground grid and a novel three transistor read and write operation to improve area density, sense margin, write performance, and write energy consumption. Despite the cell’s three transistors, the improved array configuration reduces the cell area by 22% as compared with the 2T 2MTJ cell, making it only 55% larger than a 1T 1MTJ cell. The novel mismatch tolerant read operation uses all three transistors and increases the sense margin by up to 88%. The novel variation resilient write operation also uses all three transistors and takes advantage of the inherent MTJ characteristics and complementary operation of the cell. This increases the write performance by $2\times $ and reduces the write energy by $3\times $ compared with the 2T 2MTJ cell and by $1.5\times $ compared with the 1T 1MTJ cell. [ABSTRACT FROM PUBLISHER]

We investigate the impact of wire geometry on the resistance, capacitance, and RC delay of Cu/low- $k$ damascene interconnects for fixed line-to-line pitch. The resistance is computed by applying a semiempirical resistivity model, calibrated to Cu damascene wires, integrated with a Ru-based liner, currently investigated for the 7 nm logic technology node. The capacitance is simulated by means of a 2D field solver (Raphael) by Synopsys. The impact of line dimensions is analyzed for the case of 32 nm pitch interconnects, which are representative of the 7 nm logic technology node. We show that for aspect ratios greater than 1, the resistance is more sensitive to variations of the line width rather than of the line height, because of the higher surface scattering induced by the sidewall interfaces, which are closer to each other compared with the top and bottom interfaces. For capacitance, low- $k$ sidewall damage exacerbates capacitance sensitivity to line dimensions and, for typical interconnect schemes, the impact of line width variations dominates over variations of the line height. We demonstrate that for a given pitch and dielectric stack height, the RC delay can be significantly reduced by targeting wider and deeper damascene trenches, that is, by trading capacitance for resistance, and that an optimal wire geometry for RC delay minimization exists. In addition, we show that a given RC delay can be achieved with several geometries and, therefore, $R$ and $C$ pairs, which represents a useful degree of freedom for designers to optimize system-level performance. As an application, we analyze a possible 7 nm technology scenario and show that wide and deep damascene trenches can mitigate the impact of the increased wire resistance on circuit delay. [ABSTRACT FROM AUTHOR]

For sub-7-nm technology nodes, the gate-all-around (GAA) nanowire-based device structure is a strong candidate to sustain scaling according to Moore’s Law. For the first time, the performance of two GAA device options—lateral FET (LFET) and vertical FET (VFET)—is benchmarked and analyzed at the system level using an ARM core processor, based on realistic compact device models at the 5-nm technology node. Tradeoffs among energy, frequency, leakage, and area are evaluated by a multi- Vth optimization flow. A variety of relevant device configurations, including various number of fins, nanowires, and nanowire stacks, are explored. The results demonstrate that an LFET GAA core has a larger maximum frequency than its VFET counterpart because the channel stress that can be created in the LFETs results in a larger ON current. For fast timing targets, the LFET cores are therefore superior. However, for slow timing targets (e.g., 5 ns), the VFET cores with three nanowires offer a 7% area reduction and a 20% energy saving compared with the LFET cores with 2fin/2stack at the same leakage power. [ABSTRACT FROM PUBLISHER]

Based on realistic circuit- and system-level simulations, graphene interconnects are analyzed in terms of multiple material properties, such as the mean free path (MFP), the contact resistance, and the edge roughness. The benchmarking results indicate that the advantage of using graphene interconnects occurs only under certain circumstances. The device-level parameters, including the supply and threshold voltages, and the circuit-level parameters, including the wire length and width, have large impacts on both the delay and energy-delay product (EDP). At the circuit level, one representative circuit, a 32-bit adder, is investigated, where up to 40% and 70% improvements in delay and EDP are observed. At the system-level analysis, an ARM Cortex-M0 processor is synthesized, and placement and routing are performed. After replacing copper interconnects with multilayer graphene interconnects, up to 15% and 22% performance improvements in clock frequency and EDP have been observed. It is also demonstrated that the benefits of using graphene for the ARM core processor are strongly dependent on the quality of the graphene, such as the MFP and the edge roughness. [ABSTRACT FROM AUTHOR]

In this paper, we compare the performances of FinFETs, lateral gate-all-around (GAA) FETs, and vertical GAAFETs (VFETs) at 7-nm node dimensions and beyond. Comparison is done at ring oscillator level accounting not only for front-end of line devices but also for interconnects. It is demonstrated that FinFETs fail to maintain the performance at scaled dimensions, while VFETs demonstrate good scalability and eventually outperform lateral devices both in speed and power consumption. Lateral GAAFETs show better scalability with respect to FinFETs but still consume 35% more energy per switch than VFETs if made under 5-nm node design rules. [ABSTRACT FROM AUTHOR]